Banana Leaf Disease Classification System Using Convolutional Neural Networks

Título: Banana Leaf Disease Classification System Using Convolutional Neural Networks

Autores: Gustavo de Oliveira Carvalho, Manoel Messias Silva Júnior & Matheus Pereira Alves

Resumo: This article presents the development of a disease classification system for banana plantations using Convolutional Neural Networks (CNN). The main objective is to assist producers in the early identification of diseases such as Sigatoka, Cordana, and Pestalotiopsis, which compromise production and cause financial losses. The methodology involves processing images of banana leaves to enable the classification of crop conditions, using a dataset extracted from the Banana Leaf Spot Diseases (BananaLSD) dataset. The proposed model was trained and validated using various hyperparameters to optimize disease detection performance. Performance evaluation used confusion matrices alongside precision, recall, and F1 score metrics to assess the performance of the designed classification system. The best result was obtained by combining CNN with the application of the SVM post-processing technique, achieving an overall accuracy of 89%, considering the classes cordana, healthy, pestalotiopsis, and sigatoka.

Palavras-chave: Convolutional Neural Networks; Disease Identification; Processing Images; BananaLSD; Banana Plantations.

Páginas: 7

Código DOI: 10.21528/CBIC2025-1174292

Artigo em PDF: CBIC_2025_paper1174292.pdf

Arquivo BibTeX:
CBIC_2025_1174292.bib